An Improved Anisotropic Diffusion Model For Denoising Method

The process of image acquisition, transmission, scanning can cause image quality declines. For example, the relative motion of atmospheric turbulence, the poor image in remote sensing of the optical system, etc. These can produce various types of distortion. So, it caused blurring or distortion in the image. Along with the popularization of network in modern society, the information to people is more and more important.So the information in image is the most important. So denoising becomes one of the major issues in mathematics.The classic image denoising methods have wiener filtering, inverse filtering, least-square filtering,etc. The common defects of these methods are they only as the image as a data matrix. Therefore the denoising effect is not ideal. Partial differential equations of the variational method in image processing applications more and more widely in recent years. Rudin and Osher proposed overall variational (TV, Total Variation) image model in 1992. J.weickert established anisotropic diffusion of partial differential equation model on this basis. In this paper,we fit a bivariate quadratic polynomial and pretreatment the model using Gaussian filter by scanning each sample point of the model. Then we add a determination matrix and establish a new judge rules on the basis of J. Weickert's anisotropic diffusion mode. The method maintains the original features in the small gradient and filters in large gradient.So, to some extent this paper optimizes the J. Weickert's model.